Author: Gorden Knezović, Mreža
Content Digitization/Capture transforms large volumes of incoming documents and content into classified, indexed, and searchable data, simplifying their use in a digital environment.
Content Management solutions facilitate working with documents and information while ensuring access control and regulatory compliance. This supports more straightforward decision-making and a shift toward paperless operations. We discussed the complexity of business digitalization processes and the role of artificial intelligence (AI) in this transformation with Zvonimir Mavretić, Director of the BPM International Delivery Department at ASEE.

Managed Documents
What is the core focus of your international department?
In Business Process Management (BPM), our primary focus is on content services, encompassing all aspects of document-related processes. Our ultimate goal is to assist business users by reducing costs, increasing productivity, and implementing more agile workflows tailored to the needs of modern enterprises. We accelerate digital transformation by establishing automated and efficiently managed business processes. For document-related work, we emphasize digital archiving and collaboration.
One of our solutions for digital management of meetings and formal assemblies supports the preparation and execution of sessions for various boards, from executive to supervisory boards and even governmental sessions. This results in more efficient decision-making based on relevant information with automated decision-making processes. Our digital signature solution eliminates the need for physical documents in business processes altogether.
Which documents do you address through digitization?
Let me provide a few examples. Companies track figures and related information in accounting and bookkeeping through ERP (Enterprise Resource Planning) systems. Our work within ERP involves processing associated documentation, whether physical or digital, for tasks such as invoice processing, contract management, and storing receipts, dispatch notes, and other documents. Additionally, this includes generating and managing outgoing invoices created within ERP systems.
Integration with ERP
For instance, if an invoice is entered into an ERP system digitally, we process it directly. However, we extract all necessary data for paper or PDF invoices, converting unstructured content into structured data for ERP use. Furthermore, we offer solutions for legally compliant document storage, ensuring that documents remain unaltered after entering the system—an essential requirement for documents with potential legal implications.
When a document arrives, we perform automated data extraction, significantly aided by AI. The key is to identify relevant information within the document, which is then forwarded to a Document Management System (DMS) that enables collaborative work on the documents. Working with documents has been a cornerstone of my career, and the software we discuss has been evolving for the past 20 years.
Croatia is introducing legislation requiring all business invoices to be electronic, not just those involving state enterprises or public procurement. How are your solutions aligned with this?
One long-term project we’ve undertaken is implementing a DMS that includes processing incoming invoices, sales, and purchase agreements for the entire ASEE group—not just ASEE Croatia—using our proprietary software.
In Croatia, e-invoices are currently mandatory only for transactions with the government, while in Serbia, they’re required for all entities. Romania is piloting mandatory online invoices from early 2024, while Poland plans to implement this by January 2025. These are just a few examples of our implementations, but the entire European Union is moving toward mandatory e-invoicing.
Digitalization will significantly streamline business processes as more documents become accessible online. However, document recognition will still be needed, such as for invoices from different countries or travel-related expenses, which will remain outside e-invoice systems for some time. Our goal is to ensure that our document management software can process and handle all types of documents effectively.
How do you address regulatory differences in countries where ASEE solutions are implemented?
We fully adapt our solutions to the needs of business users while strictly adhering to the regulatory requirements of each country. For instance, we recently developed solutions for North Macedonia, which required support for the Cyrillic alphabet. When creating data recognition solutions, we receive example documents from the respective country.
Our experts then define the key data to be recognized and establish the necessary rules. Once this process is complete, our system can handle any document from that country. We also tailor our Optical Character Recognition (OCR) system to successfully recognize the specific script and language of the target market.
How extensively do you use artificial intelligence in document processing?
In essence, we transform unstructured information, such as scanned documents, into structured data with the help of AI. Our solution classifies and processes physical or digital documents to maximize workflow efficiency.
Croatian Enterprises
How does the level of digitalization in Croatian enterprises compare to those in Southeast Europe?
Digitalization largely depends on regulations and the nature of a company’s operations and size. Banks, for instance, have extensively digitized their client-facing processes but still have significant room to improve internal workflows. We work with many large companies, including some in the food sector, which have been highly digitized internally.
Some of these companies began their digitalization journey as early as 2005 or 2006, with most processes digitized over a decade ago due to business needs. Digitalization has reduced the labor required for repetitive tasks and accelerated business processes.
Mid-sized companies are typically still in the early stages of digitalization, primarily using ERP systems. These businesses face greater challenges organizing the digitalization process compared to larger companies. Implementing a DMS is often more difficult due to their less structured organizational setups. For such companies, digitalization often serves as a tool for organizational restructuring.
Small businesses mostly opt for cloud computing, obtaining necessary solutions to address digitalization challenges. However, this approach often replaces the traditional model with a digital one without optimizing processes—essentially substituting paper with digital formats.
In Croatia, digitalization levels vary significantly among companies. While the general state is commendable, there are notable differences across regions, business types, and the degree of mandatory digitalization. Membership in international groups also influences the level of digitalization.
The situation is most challenging in smaller countries like Montenegro, North Macedonia, and Bosnia and Herzegovina. Serbia is comparable to Croatia in large companies due to sufficient funding for digitalization. In Romania, the IT sector is highly localized and well-developed. Slovenia, despite having fewer large companies, boasts impressive digitalization levels.
Machine Learning and AI
We use machine learning to identify document types and determine whether data recognition is necessary. This approach is particularly beneficial for complex documents or those written in scripts with unique characters. For example, we recently worked on digitizing Romanian payroll slips from the 1980s and 1990s for pension calculations. In such cases, we combine standard data recognition methods with AI tools to ensure higher-quality results.
Simpler documents are processed locally, while more complex ones undergo additional analysis using AI methods. Advanced AI techniques are now much more effective at recognizing data but require a robust infrastructure—from document preparation to processing.
What is the role of generative AI in business digitalization, and how is it applied?
There are several use cases in document management. The first involves document search, where large language models (LLMs) generate summaries, making it significantly easier to locate relevant documents. Summaries and content indexing also allow us to identify terms that traditional keyword-based searches might miss. This analysis focuses on the content’s meaning rather than exact text matches, functioning similarly to Windows Recall.
Application of Generative AI
The second application of generative AI (Gen AI) is document creation. For example, using AI, we can transcribe a meeting recording (including Croatian language support) and generate a draft of the meeting minutes. We can automatically create decision proposals from the minutes and materials related to agenda items.
AI offers several key advantages: it significantly speeds up the creation of meeting minutes and references existing organizational documents, effectively preventing “hai hallucinations.” Since we work with fixed data sets, hallucinations have not been an issue for us. In such scenarios, generative AI proves highly practical and valuable.
The key to applying AI, including generative AI, lies in identifying the right business case. AI is a powerful tool, similar to databases, and its successful application depends on recognizing where it can add value in business processes. This requires deep domain knowledge.
This interview was originally published in Croatian on Mreža.